Professional Documents
Culture Documents
Kim J. R. Rasmussen1
Abstract: The paper addresses the design of angle section columns whose legs are slender and thus subject to local buckling in their
Downloaded from ascelibrary.org by "National Institute of Technology, Karnataka" on 03/02/22. Copyright ASCE. For personal use only; all rights reserved.
ultimate limit state. For such slender angle sections, the local buckling mode is identical to the torsional mode and traditional design
procedures become excessively conservative because they account for the torsional 共local兲 buckling mode twice. The paper describes
design methods for slender equal-leg angles which ignore torsion in determining the overall buckling stress and use recently presented
effective width equations to accurately determine the bending capacity of angle sections, as required in the beam–column design approach.
The shift in the effective centroid resulting from local buckling is determined from the actual stress distribution, as obtained using
Stowell’s classical solution, rather than the effective cross section. The columns are assumed to be simply supported and thus allowed to
rotate about their principal axes.
DOI: 10.1061/共ASCE兲0733-9445共2005兲131:10共1553兲
CE Database subject headings: Columns; Beam columns; Buckling; Design; Limit states; Steel structures.
Compression
Downloaded from ascelibrary.org by "National Institute of Technology, Karnataka" on 03/02/22. Copyright ASCE. For personal use only; all rights reserved.
冦冋 冧
the effective centroid to produce an accurate design procedure for 2
关0.658c 兴Fy for c 艋 1.5
册
slender angles.
Fn = 0.877 共5兲
Fy for c ⬎ 1.5
2c
Elastic Buckling Loads for Equal-Leg Angles
Compression c = 冑 Fy
Fe
共6兲
According to classical theory 共Timoshenko 1945; Chajes and In Eq. 共6兲, Fy⫽yield stress and Fe = min兵Fext , ey其⫽lower of the
Winter 1965兲, the flexural–torsional buckling stress of a simply flexural–torsional buckling stress given by Eq. 共1兲 and the minor
supported singly symmetric column is given by axis flexural buckling stress
共ex + et兲 ± 冑
共ex + et兲2 − 4exet 1 − 冉 r20
x20
+ x20
冊 ey =
2EIy
A共KyL兲2
. 共7兲
冉 冊
Fext = 共1兲
x20
2 1−
r20 + x20 Bending
1. The flexural–torsional buckling moment for an equal-leg
where ex and et⫽major axis flexural and torsional buckling
angle subjected to uniform bending with compression at the
stresses, respectively; x0⫽distance from the centroid to the shear
free edge of the flanges is given by
center; and r0⫽polar radius of gyration. As shown in Rasmussen
共2003兲, the torsional buckling stress 共et兲 of an equal-leg angle M n = S cF c 共8兲
section reduces exactly to the local buckling stress
where Sc⫽elastic section modulus of the effective cross sec-
Fcr =
k2E t
12共1 − 2兲 B
冉冊 2
共2兲
tion corresponding to bending; and Fc⫽inelastic flexural–
torsional buckling stress determined as
冦 冧
6共1 − 兲 Fy for b 艋 0.6
冋 冉 冊册
k= 共3兲
2 102b
1.11 1 − Fy for 0.6 ⬍ b 艋 1.336
where E⫽elastic modulus; ⫽Poisson ratio; t⫽thickness; and Fc = 36
B⫽共flat兲 plate width. The plate buckling coefficient 共k兲 given by Fy
Eq. 共3兲 is the asymptotic value obtained for a length-to-width for b ⬎ 1.336
2b
ratio approaching infinity. It takes the value of 0.43 for a Pois-
son’s ratio of 0.3. 共9兲
Bending
For slender equal-leg angles bent in the plane of symmetry, it can
b = 冑 Fy
Feb
共10兲
be shown 共Rasmussen 2003兲 that: In Eq. 共10兲, Feb is the elastic flexural–torsional buckling
1. when bending causes compression at the free edges of the stress for bending about the y axis.
flanges, the flexural–torsional buckling stress becomes the In determining the effective section modulus 共Sc兲 of an
torsional buckling stress for bending which is twice the tor- equal-leg angle bent in the plane of symmetry, the unstiff-
sional buckling stress for uniform compression and ened flanges have tension at one edge and compression at the
2. when bending causes compression at the corner, the flexural– other. For such elements, the current provisions of the NAS
torsional buckling stress is so high that the section will fail Specification and AS/NZS4600 require the effective width to
再 冎
as a beam–column subjected to equal and opposite end moments
共M y兲, 共uniform bending兲. Omitting resistance factors, the interac- 1 for p 艋 0.673
tion equation is = 共20兲
共1 − 兲共1 − 0.22/ p兲/ p + for p ⬎ 0.673
P Cmy M y for 艌 1
+ =1 共16兲
Pn M ny␣y
=1
where Cmy = 1 for uniform bending and the amplification factor
where is given by Eq. 共19兲, and p = 冑 f / Fcr is determined for
␣y = 1 − P / PEy involves the Euler load for flexural buckling about
f = f 1 and with Fcr calculated from Eq. 共2兲 with k given by
the y axis. Eqs. 共12兲 and 共16兲 lead to a quadratic equation for P.
k = 1.70 + 5 + 17.12 共21兲
New Effective Width Equations for Unstiffened
Elements under Stress Gradient
Shift of Effective Centroid
General According to the NAS Specification and AS/NZS4600, the eccen-
As mentioned in the previous section, the current provisions of tricity 共ee兲 arising from the shift of the effective centroid is to be
the NAS Specification and AS/NZS4600 treat unstiffened ele- calculated as the distance between the centroids of the gross and
ments under stress gradient as if subjected to uniform compres- effective cross sections. For an equal-leg angle with sharp cor-
sion. Accordingly, the plate buckling coefficient is to be taken as ners, this leads to the following expression:
The angles tested by Popovic et al. 共1999兲 were cold rolled and
where be is obtained from the Winter equation using k = 0.43 in
in-line galvanized to produce a nominal yield stress value of 350
determining Fcr. It is implicit here that the line of action of the
force in the locally buckled stated is assumed to be located at the MPa. The measured dimensions and mechanical properties are
center of the effective width of the legs. However, while the ef- shown in Table 1. Because of the cold rolling, the yield stress
fective width equations provide an accurate estimate of the varied across the cross section with the highest values obtained at
strength of unstiffened elements, they do not accurately predict the corners. The measured yield stress value shown in Table 1
the location of the line of action of the force 共Young and Rasmus- was obtained from the middle of the flat part of the legs.
sen 1999兲. For an angle section, the shift of the effective centroid Seven fixed-ended columns and ten pin-ended columns were
共ee兲 is overestimated when using Eq. 共22兲. tested. The Le / ry values ranged from 7 to 130 and from 46 to 130
It is difficult to accurately calculate the location of the internal for the fixed-ended and pin-ended tests, respectively. For the pin-
force of a general cross section because it involves the stress ended columns, the effective length 共Le兲 is the sum of the speci-
distribution in the postbuckling range which, in turn, requires a men length and the total lengthwise dimension of the end bear-
solution of the nonlinear von Karman equations. However, in the ings. The effective length 共Le兲 is taken as half of the specimen
case of an equal-leg angle in compression, the condition at the length for the fixed-ended specimens. The use of an effective
corner is exactly that of a simple support, and the stress distribu- length of half of the specimen length accounts correctly for the
tion can therefore be found from Stowell’s 共Stowell 1949兲 classi- effect of fixed ends at long lengths, at which the column fails by
cal solution. As shown in Appendix II of Rasmussen 共2003兲, an minor axis flexural buckling. However, it is less meaningful to
accurate estimate of the shift of the effective centroid of an equal- use an effective length of half of the specimen length at short
leg angle can be obtained as lengths because the critical overall mode becomes the flexural–
torsional mode, for which the torsional component is independent
冦 冧
0 for py 艋 1.22 of the length.
The pin-ended columns were loaded with a nominal eccentric-
e0 = 5 py − 1.22 共23兲
for py ⬎ 1.22 ity of 1 / 1 , 000 of the column length relative to the gross section
16冑2 py − 0.22 centroid. Two tests were performed at each length: one where the
load was applied eccentrically towards the free edges of the
where py⫽plate slenderness calculated at the yield stress; py
flanges 共causing increased compression at the free edges兲 and one
= 冑 Fy / Fcr. The shift of the effective centroid is denoted by e0 in
where the load was applied eccentrically towards the corner. Con-
Eq. 共23兲, as distinct from the eccentricity ee which is based on the
sistently, the lowest test strength was obtained when the eccen-
centroid of the effective cross section.
tricity caused increased compression at the free edges. The geo-
metric imperfections were measured on all specimens, as detailed
in Popovic et al. 共1996兲, indicating an average overall minor axis
Tests on Equal-Leg Angles with Slender Legs
flexural imperfection of L / 1 , 305. The test strengths and column
lengths are shown in Table 3.
Wilhoite et al. Tests The loading eccentricity was introduced in the tests not for the
Compression tests on equal-leg angles with slender legs have purpose of investigating the effect of a loading eccentricity but to
been reported by Wilhoite et al. 共1984兲 and Popovic et al. 共1999兲. account for the effect of overall geometric imperfections of a
The tests are summarized as follows: The angles tested by Wil- magnitude of a thousandth of the length. Accordingly, in the com-
hoite et al. 共1984兲 were brake pressed from high strength steel parison with design loads, the columns are treated as loaded
plates. The measured dimensions and measured value of yield through the gross section centroid.
stress for the flats 共Fy兲 are shown in Table 1, see Fig. 4 for nota-
tion. Fixed-ended stub columns and three lengths of pin-ended
long columns were tested. The lengths of the long columns were
chosen so as to produce nominal L / ry values of 60, 90, and 120.
The pin-ended columns were nominally loaded through the cen-
troid of the gross cross section. However, a small clearance was
built into the pin-ended bearings to avoid locking, and an eccen-
tricity of loading may have been induced as a result of this clear-
ance. The bearings were manufactured to a tolerance that ensured
the induced loading eccentricity would not exceed 1 / 1 , 000 of the
length of the longest columns. The test strengths and column
lengths are shown in Table 2, as obtained from Fig. 18 of Wilhoite
et al. 共1984兲. Fig. 4. Nomenclature
7 1,636 120.2 59.2 0.317 using the full area rather the effective area. However, this ap-
proach would lead to conservative strengths at short lengths, since
it would not incorporate the postlocal buckling strength of unstiff-
Proposed Design Procedure for Angles with Slender ened elements. A more efficient approach is therefore to ignore
Legs the torsional mode in determining Pn and M n, and to use the
effective area in calculating Pn. This implies that the elastic buck-
General Design Approach ling stress 共Fe兲 shall be taken as the minor axis flexural buckling
stress 共ey兲, and the bending capacity becomes the section capac-
In the following development of a suitable design model for
angles with slender legs, the focus is on the torsional buckling ity determined according to Eq. 共11兲. All design models investi-
mode, the effective width calculation of angles in bending, and gated here onwards will be based on this approach.
the eccentricity required to account for the effect of the shift in The new design models 共P2 , P3 , … , P9兲 are defined in Table 4.
the effective centroid. The design strength calculations are based The main features of the P2–P5 models are that 共1兲 they ignore
on the measured cross-section dimensions and mechanical prop- torsional buckling in determining the buckling strength 共Pn兲 and
erties given in Table 1, and account for the roundedness of the the bending capacity 共M n兲 and 共2兲 the section modulus 共Se兲 is
corners. based on effective widths determined using a buckling coefficient
The design strength obtained using the current provisions of of k = 0.43, which is the current conservative approach of the NAS
the NAS Specification and AS/NZS4600 are compared with the Specification and AS/NZS4600 for unstiffened elements under
test results detailed in Tables 2 and 3 in Figs. 5共a and b兲 for the stress gradient. Various loading eccentricities 共e兲 are considered,
Wilhoite et al. 共1984兲 and Popovic et al. 共1999兲 angles, respec- including e = ee + eL, e = ee, e = eL, and e = 0. The design strengths
tively. The design strength is shown as the P1 curve in both fig-
are compared with the Wilhoite et al., and Popovic et al. test
ures. It considers the minor axis flexural and the flexural–
strengths in Figs. 5共a and b兲, respectively.
torsional buckling modes in calculating the elastic column
The design model P2 is the same as P1 except that torsional
buckling stress Fe = min兵ey , Fext其, and determines the bending
strength 共M n兲 based on the flexural–torsional buckling capacity buckling is ignored in determining Pn and M n. It can be seen from
according to Eqs. 共8兲–共10兲. The section modulus 共Se兲 is based on Figs. 5共a and b兲 that a lower strength curve in fact results from
an effective section calculated using k = 0.43. Since the columns ignoring torsion, which is because even though the axial capacity
were treated as loaded through the gross section centroid 共Pn兲 is substantially enhanced, the shift of the effective centroid
共e P = 0兲, the eccentricity is calculated according to Eq. 共14兲 as 共ee兲 is also increased, as it is calculated at the enhanced buckling
ee + eL, where ee is the distance between the centroids of the gross stress, and the combined effect is such that the interaction Eq.
Table 3. Test Strengths Reported by Popovic et al. 共1999兲; Py = 88.7 kN, ry = 9.94 mm
Tests Test number Le 共mm兲a Le / r y Pu 共kN兲 Pu / P y
Pin ended 1 459 46.2 41.7 0.466
2 458 46.1 47.2 0.528
3 676 68.0 35.2 0.394
4 676 68.0 40.1 0.448
5 862 86.7 30.9 0.345
6 863 86.8 47.5 0.531
7 1,088 109.5 25.1 0.280
8 1,088 109.5 32.1 0.359
9 1,285 129.3 17.7 0.198
10 1,286 129.4 24.7 0.276
Fixed ended 1 75 7.5 71.4 0.798
2 275 27.7 54.0 0.604
3 485 48.8 41.5 0.464
4 690 69.4 37.0 0.414
5 874 87.9 31.3 0.350
6 1,100 110.6 26.4 0.295
7 1,299 130.7 22.3 0.249
a
Le includes the end bearing dimensions for the pin-ended tests, and is half the specimen length for the fixed-ended tests.
冦 冧
共18兲 are used for determining M n, and the shift of the effective 1 for py 艋 1.22
centroid 共e0兲 is calculated using Eq. 共23兲. Equal-leg angle col- = 0.68 共28兲
umns with slender legs can thus be designed using a beam– for py ⬎ 1.22
共py − 1兲1/4
column approach with the eccentricity calculated as
Eq. 共26兲 predicts the capacity of a short length of an equal-leg
e = e0 + e P 共24兲 angle loaded through the gross section centroid. A simple column
where e0 is calculated according to Eq. 共23兲 and the loading ec-
centricity 共e P兲 is measured from the centroid of the gross cross
section, as shown in Fig. 1共d兲. It is not necessary to include the
additional eccentricity eL = L / 1 , 000 currently specified in the
American and Australian standards.
As shown in Fig. 6共b兲, the design model P9 also produces
accurate strength predictions for the Popovic et al. 共1999兲 fixed-
ended tests, except at short lengths where the tests strengths are
significantly higher than the design curve because the shift in the
effective centroid does not induce overall bending in fixed-ended
columns 共Rasmussen and Hancock 1993兲. In fact, it could have
been expected that the fixed-ended test points would lie close to
the P5 pure column design curve. However, the P5 curve has been
calibrated for columns failing by flexural buckling while slender
angle section columns fail by flexural–torsional buckling at short
and intermediate lengths. Furthermore, it is not clear whether the
effective length concept is readily applicable to columns failing
primarily by torsion at short lengths. The comparison shown in
Fig. 6共b兲 demonstrates that fixed-ended slender angle section col-
umns cannot simply be designed as concentrically loaded col-
umns using the strength curve for flexural buckling. The design
model P9 works well for the Popovic et al. 共1999兲 section but
becomes conservative for fixed-ended equal-leg angle columns
with more slender cross sections.
follows from the figures that the column design strength Pn,r is in Gregory Hancock for his comments to the manuscript.
close agreement with the beam–column design model proposed in
this paper, and is in good agreement with the test strengths. References